Dolby Elevates the Quality of Lossless Audio on Blu-ray - Page 2

The initial advantage of quad def to studios is that it cannot be streamed, thus opening up a whole new high-profit market.

Is it an objective improvement or a euphonic effect?

Also, how does one remove artifacts that were caused by the A-to-D process, as the Dolby press material claims to do? Specifically, if the A-to-D process sampled at 48 kHz, there always has to be a brickwall filter with a stopband at 24 kHz. How do you undo that?


Both papers work out the math very well, but neither really speak to the audibility or the audible effects of such filters. The Dunn paper conjectures on it, but it's really about the math of the process.

Let me ask a basic question: at what frequency is the pre-ringing occurring? The Dolby press material shows two graphs, one with pre-ringing, and one without, but the X-axis has no units, so one cannot work out the frequency of the ringing.

Assuming you have the answer to the first question, consider a second question: if the frequency of the ringing is above human audibility (say >20kHz), is it still important?

Consider a third question: if in order to create an impulse response with no preringing, one must create a phase shift (and the math shows that one must), then when is that phase shift audible, and what are its effects?

Some people have tried to answer those questions, and the answers were, to me, surprising.

I don't want my disc sounding "better" than the master it came from I want it sounding the same.

Dolby belonged dead, when two products are the same- one must die.

This Technical analysis of the Meridian Apodizing filter by someone who writes unofficial Meridian firmware to improve on their apodizer filter is an interesting read. Note:





So this is not entirely a new invention.

The same author's Meridian Apodizing filter: a review indicates the Meridian filter is not without disadvantages. The last paragraph reflects my sentiment already expressed before, that it is about choice for the consumer:


Linn, another high-end audio company does not use apodizing filter in their upsampling products. Clearly, not everyone thinks apodizing filter is the only solution or necessarily the best, or a must-have.

Maybe further technical stuff should go to the audio theory forum where there are more people from the other side.

By upsampling, the stop-band is an octave higher, allowing room to use a gentler filter.


You had asked for info on the science and engineering.


According to Bob Stuart: >>the filter is going to ring at something below the half-sample rate, which puts
the ringing in the audioband with a 44 kHz sampling rate<< More insights about it in the TAS article.

I guess the auto industry didn't get the memo.

TrueHD was first, on time and working. So by your proclamation...

Don't forget, smaller file size too.

Yes, but before you can upsample, you need to lowpass-filter the original signal at half its original sample rate so the upsampling won't create aliasing. Traditionally, this is done with a brickwall filter which leads to the pre-ringing. This means that upsampling a 48 kHz signal requires some kind of lowpass filter below 24 kHz.

Meridian and other advocates of apodising do it with a minimum phase filter which has one of two effects: either you have drooping amplitude response at the top of your passband with some phase shift or you have a lot of phase shift in your passband with a flat amplitude response. They are simulating a very high quality, steep-slope analog filter response in DSP.

It seems Meridian has chosen phase shift so the amplitude response is flat to 20 kHz (as can be deduced from the Stereophile measurements of the 808 CD player). It seems to me that this phase shift is a more likely explanation for any audible changes than suprasonic (>20kHz) signals.


That's one part. Another part is the audibility of such filters.


So if it rings at 22 kHz, does it matter when almost everyone's hearing only goes up to 20 kHz (if they're lucky)?

Even worse, at 48 kHz, the ringing will be closer to 24 kHz. Even if you assume that humans can hear above 20 kHz, how many speakers have linear response above 20 kHz?

The source content has already been bandlimited <24 kHz by the original brickwall filter.


But the new filter is being simulated at the 96 kHz sample rate, not 48 kHz, to deal with upsampling image components above the 48 kHz audio band as opposed to 24 kHz as before.


The ringing is not the superposition of a 22 kHz sine wave tone into the audio, but a ringing in the time domain (impulse response) that causes pre-echoes of the audio at frequencies that extend well below, diminishing as frequency drops.

Mi4? Dotm?

I didn't mean it mean. Just saying since the majors all chose DTS's codec, Dolby had to choose a gimmick to remain relevant.

Simpler tools aren't a gimmick since that's why they chose DTS-MA to begin with.

Disney, Sony, Paramount, and Warner all initially chose TrueHD until these new simplified authoring tools were available.

Can anyone in the industry explain how its "easier"?

HDMA encoding includes the lossy core inside. If the lossless decode is bit perfect to the source (which is easy to check), then the lossy core is also correct. One QC pass.

TrueHD's companion lossy "core" is not used to reconstruct the lossless audio, so it has to be QC'd separately. 2 QC passes. That's extra time and that's extra cost.

It would be possible to automate this process to mitigate any such differences, but in the early days that was not part of the program.

Didn't know truehd had a lossy core. I thought it was just downcoverted to 2 channel :d

I used the term "core" so as to describe it in similar terms as the lossy core in HDMA. But it is actually just a companion that travels on the disc exactly the same way as the DTS core. The difference is that while the HDMA decoder has to decode both the core and the lossless streams in parallel, the TrueHD decoder only decodes the lossless stream.

This would be really sad if true

It will be interesting to see how many people can hear an appreciable improvement with this over the prior 48hz on a movie soundtrack. It would seem music would be more likely to appreciate this rather than movies. Also, would most speakers be able to transmit the improved upsampling to audibly percieve it or would this require more audiophile quality speakers?

Except it is not.

Yes, but the sampling process creates a signal that has images at every integer multiple of Fs. All the original brickwall filter does is to limit bandwidth so that those images do not alias (overlap). If you upsample, you must bandlimit the digital samples to less than half your new sampling rate. Most people will bandlimit it to half the original sampling rate, but technically you could just do half the new sampling rate.


I don't believe the math bears this out, and I don't think either of the Dunn or Craven papers show this either. Convolve a sinc function (sin(x)/x or the time domain dual of a brickwall filter) with a properly sampled signal, and you will get back the sampled signal that was bandlimited. The overshoots will appear only at the top frequency of the brickwall filter.

If what you say were true, we would be measuring massive THD+N in digital players using linear phase filters.

That's exactly what the new apodizing filter does.


Massive is a subjective term. THD+N is measured with steady state sine waves, not impulses.

First of all, they're not quite the same, even though they have the same purpose and theoretically, since they're both "lossless" should sound the same. Secondly, if you give someone a monopoly, they abuse it.

I have to express skepticism here. As with many other "I heard a difference" audio issues, the question arises "was the comparison made under proper double blind conditions"? It seems to me that the experiences with SACD and DVD-A don't show a significant difference, else those formats would have succeeded.

Significant is a subjective term. Are you saying there was no discernible difference between CD and DVD-A 96 kHz? I do not see that as being the consensus, and we find music discs still being produced at 96 kHz to this day. The failure of DVD-A and SACD do not rest on this question -- there were many other factors that led to their demise having nothing to do with sound quality.

Furthermore, unlike SACD (in particular) and possibly DVD-A, the usual mode of listening is directly to analog through their internal, presumably high quality DACs. That's very different than what is being proposed here for Blu-ray upsampling, where the PCM will be passed forward and subjected to further signal processing, and DACs of unknown quality. Meridian found that it makes a difference where the upsampling is done, at the source vs at the DACs. Additional benefits accrue when it is done at the source, as that spreads the quantization byproducts of all the downstream processes over the wider bandwidth, even though it multiplies the processing burden for the DSPs. Neither of these factors are likely to be major issues in a Meridian surround processor like the 861, but the quantization benefits could be more important in normal AVRs. If they can handle 96 kHz properly, that is.

It should be pointed out that there may be AVRs that are unable to reveal the full benefits of these upsampled discs, if for example they internally downsample to 48 kHz to lighten the load on the DSPs under certain conditions. This has been the case for heavy algorithms like Audyssey, for example, in the early days when it could barely fit. Newer AVRs have more powerful DSPs, sometimes 2 of them, so they should be fine staying 96 kHz all the way through.

Unfortunately, the AVR makers will not tell you if their products downsample when the going gets rough. It is of course easy to find out with an appropriate test disc and measuring tools.

I'm saying that there was an extensive double-blind study comparing SACD and DVD-A with the identical recording passed through a 16/44.1 ADA chain, and the results weren't statistically distinguishable from chance. Dolby would make a convincing case for this product if it would publish the results of a DBT showing that people CAN hear the difference.

The failure to detect a difference does not mean there is none. And since the use case was different than the one used for BD movies (as I mentioned), it's apples vs. oranges.


I doubt anyone would believe a test sponsored by the proponent.

People don't have to prove a negative ("there isn't a difference"), nor does it make sense for them to try to do so. It's the advocate for the technology that needs to show people hear a difference under controlled (ie double blind) conditions.


People might be skeptical, but if the double blind methodology was thoroughly laid out and appeared to be sound, that would make for a strong case for the technology, much more powerful than testimonials or descriptions of it.

They did that with DD/DTS and that didn't stop people from proclaiming DTS was automatically better no matter what, even though the blind test showed even at 448kbps vs 1536kbps they couldn't readily tell the difference between the two.

Dts is way better.... lol jk